KR100468844B1 - Optimal scanning method for transform coefficients in image and video coding/decoding - Google Patents

Abstract

An optimal scanning method for encoding / decoding a video signal is disclosed. An image signal encoding method using discrete cosine transform includes a selection step of selecting at least one reference block; And a scanning step of generating a scanning order of blocks to be encoded from the reference blocks, and scanning in accordance with the generated scanning order. The selecting step is a step of selecting the at least one block temporally or spatially adjacent to the block to be encoded, and the scanning step is a step of obtaining a probability of occurrence of coefficients other than "0" from the selected at least one reference blocks. ; And generating the scanning order in ascending probability of occurrence. The generating of the scanning order in ascending probability of occurrence includes generating the scanning order by a zigzag scanning method when the occurrence probability is the same. The scanning method for encoding / decoding according to the present invention has an advantage of increasing signal compression efficiency because an optimal scanning method can be used.

Description

Optimal scanning method for transform coefficients for image and video coding / decoding

The present invention relates to encoding / decoding of video signals, and more particularly, to an optimal scanning method of transform coefficients for encoding / decoding still images or moving images by increasing encoding / decoding efficiency.

In MPEG-1 (moving picture expert group phase 1), MPEG-2, MPEG-4, H.261, H.263, or JPEG, the video signals (e.g. still image or video) are discrete cosine transform. (Hereinafter referred to as "DCT"). Also, the compressed or encoded video signals are decoded through an inverse discrete cosine transform (hereinafter referred to as "IDCT"). The key point to increase the efficiency of encoding / decoding is to select and scan the optimal scanning method.

The scan pattern described in US Patent No. 5,500,678 is designed to increase coding efficiency and is selected as a scan method of MPEG-4 intra coding. The scan method described in '678 is a zigzag scan method and Use the defined scanning method.

Therefore, the zigzag scan method and the predefined scan method are not efficient in encoding all video signals. In addition, since the information on the selected scanning method should be encoded and provided to the decoder, the amount of bits transmitted to the decoder is increased. Since the scanning method of '678 is predefined, there is a limit in selecting an optimal scanning method for various encoding blocks.

In addition, the scanning method described in US Patent No. 6,263, 026 has a problem that if the number of scanning methods is too large, the selected data word is too long, thereby reducing the coding efficiency. In addition, the predefined finite number of scanning methods used in '026 are not efficient for all image signals or data.

Accordingly, a technical problem of the present invention is to increase the compression efficiency of an image signal by providing an optimal scanning method for various encoding / decoding blocks.

The detailed description of each drawing is provided in order to provide a thorough understanding of the drawings cited in the detailed description of the invention.

1 shows a block diagram of a typical DCT coding system.

2 shows a general zigzag scanning sequence.

3 shows a general vertical scanning sequence.

4 shows a general horizontal scanning sequence.

5 shows an example of the output signal quantized by FIG.

6 is a flow chart for implementing an optimal scanning method according to the present invention.

7 shows a first example of reference blocks.

8 shows a second example of reference blocks.

9 to 11 show a third example of the reference blocks.

12 to 14 show a fourth example of the reference blocks.

FIG. 15 illustrates blocks implementing step 600 of FIG. 6.

FIG. 16 illustrates a block implementing operation 610 of FIG. 6.

FIG. 17 illustrates a block implementing steps 620 and 630 of FIG. 6.

18 and 19 show scanning schemes adopted in the H.26L video encoder.

20 and 21 show a scheme for converting the scanning order of the single scan mode to the scanning order of the dual scan mode according to the present invention.

22 to 24 show an embodiment of converting the scanning order of the single scan mode to the scanning order of the dual scan mode according to the present invention.

25 is a graph showing the sum of the total lengths of the run lengths generated by the scanning method and the Foreman sequence according to the present invention.

26 is a graph showing the sum of the total lengths of run lengths generated by the scanning method and the Coastguard sequence according to the present invention.

27 is a graph showing the total length of run lengths generated by the scanning method and the Hall sequence according to the present invention.

According to an aspect of the present invention, there is provided a method of encoding an image signal through discrete cosine transform, selecting at least one reference block; And a scanning step of generating a scanning order of blocks to be encoded from the reference blocks, and scanning in accordance with the generated scanning order.

The selecting step is a step of selecting the at least one block temporally or spatially adjacent to the block to be encoded, and the scanning step is a step of obtaining a probability of occurrence of coefficients other than "0" from the selected at least one reference blocks. ; And generating the scanning order in ascending probability of occurrence.

The generating of the scanning order in ascending probability of occurrence includes generating the scanning order by a zigzag scanning method when the occurrence probability is the same.

Also, a method of encoding an image signal through discrete cosine transform may include obtaining a probability of occurrence of coefficients other than "0" from at least one selected reference block; And determining the scanning order of blocks to be encoded in the order of the high probability of occurrence, and scanning the blocks according to the scanning order.

The at least one reference blocks are blocks that are temporally or spatially adjacent to the block to be encoded. When the probability of occurrence is the same, the scanning order is determined by the zigzag scanning method. The injection is preferably a single injection or a double injection.

Also, a method of decoding an image signal through an inverse discrete cosine transform for achieving the technical problem includes selecting at least one reference block; And a scanning step of generating a scanning order of blocks to be decoded from the reference blocks and scanning in accordance with the generated scanning order.

The selecting step is a step of selecting the at least one block that is temporally or spatially adjacent to the block to be decoded. The scanning step may include obtaining a probability of occurrence of coefficients other than "0" from at least one selected reference block; And generating the scanning order in ascending probability of occurrence. The generating of the scanning order in ascending probability of generation occurs when generating the scanning order using a zigzag scanning method when the probability of occurrence is the same.

Also, a method of decoding an image signal through an inverse discrete cosine transform may include obtaining a probability of occurrence of coefficients other than "0" from at least one selected reference block; And determining the scanning order of blocks to be encoded in the order of the high probability of occurrence, and scanning the blocks according to the scanning order.

The at least one reference blocks are blocks that are temporally or spatially adjacent to the block to be decoded. When the probability of occurrence is the same, the scanning order is determined by the zigzag scanning method. Preferably, the injection is a single or double injection.

In order to fully understand the present invention, the operational advantages of the present invention, and the objects achieved by the practice of the present invention, reference should be made to the accompanying drawings which illustrate preferred embodiments of the present invention and the contents described in the accompanying drawings.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. Like reference numerals in the drawings denote like elements.

1 shows a block diagram of a typical DCT coding system. Referring to FIG. 1, a DCT coding system (or an encoder 100) may include a motion estimator 10, a subtractor 20, a DCT coder 30, and a quantizer 40. ), Variable length coder (50), rate control unit (60), inverse quantizer (dequantizer) 70, IDCT unit (inverse DCT; 75), adder 80, reconstruction buffer (frame memory or reconstruction) a buffer 85 and a motion compensator 90. The variable length encoder (hereinafter referred to as "VLC") includes a scan method selector 51 and an entropy encoder 53. Video encoding apparatuses used in MPEG-1, MPEG-2, MPEG-4, H.261, H.263, or JPEG are well known in the art and will be described briefly.

The motion evaluator 10 generates a motion vector in response to the video signal and outputs the motion vector to the subtractor 20. The subtractor 20 outputs the difference between the output signal of the motion evaluator 10 and the output signal of the motion compensator 90 to the DCT encoder 30.

The entire image is divided into m × n (where m and n are natural numbers) sample blocks, and the divided m × n sample blocks are input to the DCT encoder 30 by one block. The DCT encoder 30 converts a video signal in a spatial domain into a transform coefficient in a frequency domain. In other words, the DCT encoder 30 converts the m × n sample block into an m × n coefficient block. For example, the entire image may be divided into 4 × 4, 8 × 8, or 16 × 16 sample blocks. The quantization unit 40 quantizes the m × n coefficient block.

The inverse quantizer 70 inverse quantizes the output signal of the quantizer 40, and the IDCT unit 75 inversely DCTs the output signal of the inverse quantizer 70. The adder 80 adds the output signal of the motion compensator 90 and the output signal of the IDCT unit 75. The recovery buffer 85 stores the output signal of the adder 80. Therefore, the signal corresponding to the original video signal is decoded in the reconstruction buffer 85.

The motion compensator 90 compensates for the motion of the output signal of the reconstruction buffer 85, and the variable length encoder 50 is a value (level) generated with a high probability in response to the output signal of the quantizer 40. Short codes are used, and low-probability values (levels) are assigned long codes to reduce the total number of bits in the data flow. The variable length encoder 50 includes a scanning method selector 51 and an entropy encoder 53.

Since the scanning method selection unit 51 selects a predetermined scanning method in response to the output signal of the quantization unit 40, the two-dimensional data is converted into the one-dimensional data by the scanning method selection unit 51. The entropy encoder 53 outputs coded data (for example, a compressed bitstream) in response to the output signal of the scanning method selector 51. The entropy encoder 53 may use Hoffman coding or other coding.

The rate controller 60 controls the quantizer step size of the quantizer 40 in response to the output signal of the entropy encoder 53.

The present invention relates to the operation of the scanning method selecting unit 51. The scanning method selecting unit 51 according to the present invention selects an optimal scanning method by referring to at least one reference block that is already encoded, thereby increasing the compression efficiency. Let's do it. The operation of the scanning method selection unit 51 is described in detail with reference to FIG.

Referring to FIG. 1, an intra frame for encoding an image signal itself will be described below. The video signal (e.g., still image or video) is divided into blocks of constant size (e.g., m x n sample blocks), and each block is quantized in two dimensions by the DCT encoder 30 and the quantizer 40. When the quantized data is made into one-dimensional data through the scanning method selector 51, the entropy coding unit 53 converts the one-dimensional data into a compressed bitstream. The decoder decodes the original video signal from the bitstream as opposed to the encoding of the DCT coding system 100.

Next, an inter frame will be described with reference to FIG. 1. The inter frame uses a previous video signal to encode a current video signal, and the difference between the output signal of the motion estimation unit 10 and the output signal of the motion compensator 90 is determined by the DCT encoder 30 and the quantizer ( 40) to quantize in two dimensions. If the quantized data is made into one-dimensional data through the scanning method selector 51, the entropy coding unit 53 converts the one-dimensional data into a compressed bitstream. The decoder decodes the original video signal from the bitstream as opposed to the encoding of the DCT coding system.

2 shows a general zigzag scanning sequence. 3 shows a general vertical scanning sequence. 4 shows a general horizontal scanning sequence. The scanning patterns of Figs. 2 to 4 show examples of scanning schemes used in MPEG-4. Sample blocks of FIGS. 2 to 4 are 8x8.

5 shows an example of the output signal quantized by FIG. Referring to FIG. 5, 300 represents an 8 × 8 quantized coefficient block, and 301 represents a quantized DC coefficient, in which case the DC coefficient is five. 302 through 304 represent quantized AC coefficients, in which case the AC coefficients represent -1, 3, and 1, respectively.

In the case of MPEG-4, the scanning method selector 51 of FIG. 1 extracts information on a plurality of symbols (run, level, last) by scanning the block 300 according to the scanning method of FIGS. do. In this case, the run represents the number of "0" s between the non-zero "0" data and the data which is not the current "0" during the scanning according to the scanning method of FIGS. 2 to 4, and the level is the current "0". Non-data level (value), and last indicates the presence or absence of data "0" after the current non-zero data.

For example, in the symbols (0, 5, 0) 351, the first 0 indicates the number of "0" s in front of the DC coefficient 301 in zigzag scanning, and the second 5 indicates the current DC coefficient. A third 0 indicates that there is non-zero data after the DC coefficient. Further, in the symbols (5, 1, 1) 354, the first 5 indicates the number of "0" s in front of the AC coefficient 304 in the zigzag scanning method, and the second 1 indicates the current AC coefficient. Is 1, and the third 1 indicates that there is no non-zero data after the AC coefficient 304. In addition, the vertical scanning method may be easily understood with reference to FIGS. 3 and 5, and the horizontal scanning method may be easily understood with reference to FIGS. 4 and 5.

Each VLC (356 to 359, 367 to 370, or 378 to 381) corresponds to each bit pattern corresponding to a table of MPEG-4 entropy encoders for each symbol (351 to 354, 362 to 365, or 373 to 376). Indicates. That is, the bit pattern for the symbols (0, 5, 0) 351 is 011000 (356), and the bit pattern for the symbols (0, -1, 0) 363 is 101.

In addition, each bit amount 360, 371, or 382 represents the sum of the bits corresponding to each symbol 351 to 354, 362 to 365, or 373 to 376. For example, the four symbols of the zigzag scanning method consist of 24 bits in total, the four symbols of the vertical scan method consist of 19 bits in total, and the four symbols of the horizontal scan method consist of 30 bits in total. Since the run of each symbol is different, the amount of bits generated by each scanning method is different.

6 is a flow chart for implementing an optimal scanning method according to the present invention. Referring to FIG. 6, at least one block temporally or spatially adjacent to a block to be encoded (hereinafter, referred to as a “coding block”) is selected as reference blocks (step 600). The selected reference blocks are already coded and quantized blocks.

In operation 610, a probability of occurrence of non-zero coefficients is calculated from each reference block at each coefficient position. Step 620 orders the occurrence probability of coefficients other than "0" from descending probability to descending order. If the probability of occurrence is the same, the zigzag scanning method is ordered (step 630). In step 640, the scanning method determined in steps 620 and 630 is selected, and the coding block is scanned in the selected scanning method.

7 shows a first example of reference blocks. Referring to FIG. 7, spatially adjacent blocks 710 to 740 are selected as reference blocks in order to select a scanning method of 750 blocks. The reference blocks 710 to 740 of FIG. 7 may be used for video coding of an intra frame. The reference blocks 710 to 740 are blocks that are already encoded.

8 shows a second example of reference blocks. Referring to FIG. 8, blocks 811 to 819 that are temporally adjacent to the current frame t in the reference frame t-1 are selected to select a scanning method of the 831 block (coding block) of the current frame t. Reference blocks are selected. Reference blocks 811 to 819 of FIG. 8 may be used for video coding of an inter frame. The reference frame t-1 represents a previous frame of the current frame t.

9 to 11 show a third example of the reference blocks. The scanning scheme of the coding block of FIG. 9 is determined with reference to all adjacent blocks a, b, c and d, and the scanning scheme of the coding block of FIG. 10 refers to all adjacent blocks a, b and c. The scanning scheme of the coding block of FIG. 11 is determined with reference to all adjacent blocks a and b. Each block (a, b, c, or d) is an m × n block, which is a sample block.

12 to 14 show a fourth example of the reference blocks. The scanning method of the coding block of FIG. 12 is determined by referring to the most suitable block among the adjacent blocks a, b, c, and d, and the scanning method of the coding block of FIG. 13 is determined by the adjacent blocks (a, b, And c) is determined by referring to the most suitable one block, and the scanning scheme of the coding block of FIG. 14 is determined by referring to the most suitable one of the adjacent blocks a and b. Each block (a, b, c, or d) is m × n blocks. 7 to 14 are merely examples for selecting reference blocks.

FIG. 15 illustrates blocks implementing step 600 of FIG. 6. Referring to FIG. 15, the reference blocks 710 to 740 are blocks that are spatially adjacent to the coding block 750 as blocks that are already coded. Each block 710, 720, 730, 740 is an 8 × 8 block.

FIG. 16 illustrates a block implementing operation 610 of FIG. 6. Referring to FIGS. 15 and 16, the 8 × 8 sample block 800 is composed of 64 subblocks, and the number of each subblock represents a probability of a coefficient that is not "0" compared to the reference blocks. For example, if the DC coefficients of each of the reference blocks 710, 720, 730, and 740 are 5, 3, 3, 4, the probability of the coefficient that is not "0" compared to the reference blocks is 4/4. In addition, if the AC coefficients of each of the reference blocks 710, 720, 730, and 740 are 4, 2, 0, and 4, the probability of a coefficient other than "0" compared to the reference blocks is 3/4.

FIG. 17 illustrates a block implementing steps 620 and 630 of FIG. 6. Each numeral of FIG. 17 represents a scanning order, and the scanning order of FIG. 17 is determined in the order of occurrence probability of coefficients other than " 0 ".

The scanning order of FIG. 17 is not a predefined scanning order and is determined corresponding to at least one adjacent block in time or space. Therefore, since the scanning method of the coding block according to the present invention is the most suitable and efficient scanning method for the current coding block, the compression efficiency of the video signal increases.

18 and 19 show scanning schemes adopted in the H.26L video encoder. FIG. 18 is the same as a conventional zigzag scan in single scan mode, and FIG. 19 shows a double scan mode. Double injections are two injections that do not overlap. H.26L is being created in the International Telecommunications Union Telecommunications standardization sector (ITU-T).

20 and 21 show a scheme for converting the scanning order of the single scan mode to the scanning order of the dual scan mode according to the present invention. 20 shows a scanning sequence (a → b → c → d →, ..., → n → o → p) generated in a single scan mode according to the present invention, and FIG. 21 is a dual scan mode according to the present invention. The generated scanning sequence is shown. 20 and 21, the scanning order of the dual scan mode is the even scanning order (a → c → e → g → i → k → m → o) and the odd scanning order (b → d → f → h → j → l → n → p), respectively, in order.

22 to 24 show an embodiment of converting the scanning order of the single scan mode to the scanning order of the dual scan mode according to the present invention. FIG. 22 illustrates a probability of a non-zero coefficient generated from at least one reference block according to the present invention, FIG. 23 illustrates a scanning sequence of a single scanning mode determined according to FIG. 22, and FIG. Fig. 21 shows the scanning order of the single scan mode in which the scanning order of the single scan mode is converted in accordance with FIG.

The even scanning order (0 → 2 → 4 → 6 → 8 → 10 → 12 → 14) of FIG. 23 is converted to the order of ( 0 → 1 → 2 → 3 → 4 → 5 → 6 → 7 ) in FIG. The odd scanning order (1 → 3 → 5 → 7 → 9 → 11 → 13 → 15) in FIG. 23 is converted in the order of (0 → 1 → 2 → 3 → 4 → 5 → 6 → 7) in FIG.

25 is a graph showing the sum of the total lengths of run lengths generated by the scanning method and the Foreman sequence according to the present invention. 26 is a graph showing the sum of the total lengths of run lengths generated by the scanning method and the Coastguard sequence according to the present invention. 27 is a graph showing the total length of run lengths generated by the scanning method and the Hall sequence according to the present invention. 25 to 27 show the total length of the run length generated by applying 100 intra frames to the scanning method according to the present invention and the conventional zigzag scanning method. 25 to 17, qp represents a quantizer step size, and as the qp increases, the image quality decreases and the data amount decreases. However, as qp decreases, the amount of data increases and the image quality improves.

25 to 27, it can be seen that the run length of the scanning method according to the present invention is reduced compared to the run length of the conventional zigzag scan. Therefore, the scanning method according to the present invention has the effect of increasing the video compression efficiency.

Also, since the decoder can generate the same scanning method as the scanning method generated by the encoder according to the present invention, the decoder does not need any additional information on the scanning method used in the encoder. The scanning method for encoding according to the present invention has the effect that all possible scanning methods can be used, and since the scanning method for decoding according to the present invention can use all possible scanning methods, the scanning method for conventional encoding can be used. The compression efficiency is increased.

Although the present invention has been described with reference to one embodiment shown in the drawings, this is merely exemplary, and those skilled in the art will understand that various modifications and equivalent other embodiments are possible therefrom. Therefore, the true technical protection scope of the present invention will be defined by the technical spirit of the appended claims.

As described above, the scanning method for encoding / decoding according to the present invention has an advantage of increasing signal compression efficiency because an optimal scanning method can be used. In addition, the scanning method for encoding / decoding according to the present invention has the effect of using any combination of scanning methods.

In addition, the scanning method for encoding according to the present invention has the effect of reducing the bit generation amount by reducing the length of the run length (run length).

Claims (20)

In the method for encoding a video signal through a discrete cosine transform, Selecting at least one block temporally or spatially adjacent to a block to be encoded as a reference block; Obtaining a probability of occurrence of coefficients other than "0" of the selected at least one reference block; And And generating the even and odd scan sequences separately in the order of the high probability of occurrence, and scanning the same scan sequence so as not to overlap each other according to the generated even and odd scan sequences. Way. delete delete In the method for encoding a video signal through a discrete cosine transform, Selecting at least one block temporally or spatially adjacent to a block to be encoded as a reference block; Obtaining a probability of occurrence of coefficients other than "0" from the selected at least one reference block; And Determining the scanning order in ascending probability of occurrence, and if the probability of occurrence of the coefficients is the same, determining the scanning order by a zigzag scanning method and scanning according to the determined scanning order. . delete In the method for encoding a video signal through a discrete cosine transform, Obtaining a probability of occurrence of coefficients other than "0" from the selected at least one reference block; And And generating the even and odd scanning orders of the blocks to be encoded in the order of the highest probability of occurrence, and scanning the data in a non-overlapping manner according to the generated even and odd scan sequences. A video signal encoding method. delete In the method for encoding a video signal through a discrete cosine transform, Obtaining a probability of occurrence of coefficients other than "0" from the selected at least one reference block; And Determining the scanning order of blocks to be coded in ascending probability of occurrence, and determining the scanning order by a zigzag scanning method when the probability of occurrence of the coefficients is the same, and scanning according to the determined scanning order. A video signal encoding method. delete delete In the method for decoding an image signal through an inverse discrete cosine transform, Selecting at least one block temporally or spatially adjacent to the block to be decoded as a reference block; Obtaining a probability of occurrence of coefficients other than "0" from the selected at least one reference block; And And generating the even and odd scan orders separately in the order of the high probability of occurrence, and scanning the data in a non-overlapping manner according to the generated even and odd scan orders. Way. delete delete In the method for decoding an image signal through an inverse discrete cosine transform, Selecting at least one block temporally or spatially adjacent to the block to be decoded as a reference block; Obtaining a probability of occurrence of coefficients other than "0" from the selected at least one reference block; And Generating the scanning order in ascending order of occurrence probability, and generating the scanning order by a zigzag scanning method when the probability of occurrence of the coefficients is the same, and scanning according to the generated scanning order. Decryption method. delete A method of decoding a video signal through inverse discrete cosine transform, Obtaining a probability of occurrence of coefficients other than "0" from the selected at least one reference blocks; And And generating the even and odd scanning orders of the blocks to be decoded in the order of high probability of occurrence, and scanning the blocks without overlapping each other according to the generated even and generated odd scanning orders. A video signal decoding method. delete A method of decoding a video signal through inverse discrete cosine transform, Obtaining a probability of occurrence of coefficients other than "0" from the selected at least one reference blocks; And Determining the scanning order of blocks to be decoded in ascending probability of occurrence, and determining the scanning order by a zigzag scanning method when the probability of occurrence of the coefficients is the same, and scanning according to the determined scanning order. A video signal decoding method. delete delete